Exposure to a non-ionic surfactant induces a response akin to heat-shock apoptosis in intestinal epithelial cells: implications for excipients safety

Amphipathic, non-ionic, surfactants are widely used in pharmaceutical, food, and agricultural industry to enhance product features; as pharmaceutical excipients they are also aimed at increasing cell membrane permeability and consequently improving e.g. oral drugs absorption. Here we report on the concentration- and time-dependent succession of events occurring throughout, and subsequent exposure of Caco- 2 epithelium to a ‘typical’ non -ionic surfactant (Kolliphor® HS15) to provide a molecular explanation for non-ionic surfactant cytotoxicity. The study shows that the conditions of surfactant exposure, which increase plasma membrane fluidity and permeability, produced rapid (within 5 minutes) redox and mitochondrial effects. Apoptosis was triggered early during exposure (within 10 minutes) and relied upon an initial mitochondrial membrane hyperpolarization (5-10 minutes) as a crucial step, leading to its subsequent depolarization and caspase-3/7 activation (60 minutes). The apoptotic pathway appears to be triggered prior to substantial surfactant- induced membrane damage (observed ≥60 minutes). We hence propose that the cellular response to the model non-ionic surfactant is triggered via surfactant-induced increase in plasma membrane fluidity - a phenomenon akin to the stress response to membrane fluidization induced by heat shock - and consequent apoptosis. Therefore, the fluidization effect that confers surfactants the ability to enhance drug permeability may also be intrinsically linked to the propagation of their cytotoxicity. The reported observations have important implications for the safety of a multitude of non-ionic surfactants used in drug delivery formulations, and to other permeability enhancing compounds with similar plasma membrane fluidizing mechanisms. mitochondrial membrane potential Mitotracker Red FCCP µM) and Kolliphor HS15 exposure on ROS and CellTox Green assay results.


Introduction
Non-ionic surfactants are widely used in pharmaceutical, food, and agricultural industries. In the pharmaceutical industry, for example, they are employed to improve drug solubility, or to enhance transepithelial drug transport and hence increase its bioavailability. [1][2][3] In addition to a multitude of pre-clinical studies, non-ionic surfactants, such as the alkyl glycosides, are employed in commercially available permeability enhancing formulations, including Intravil® (Aegis Therapeutics). 4 Previous studies attribute enhancement in epithelial permeability by non-ionic surfactants to increased cell membrane fluidity 2,5 and /or to formation of channels (pores) in the plasma membrane. 6 The consequences of these changes in membrane fluidity and structure further extend to changes in function of membrane bound and cytoskeletal proteins, including Pglycoprotein and F-actin, respectively, whereby at sufficiently high concentrations, destruction of plasma membrane integrity and ultimately cell lysis could occur. [7][8][9] Apoptotic and necrotic death pathways have both been reported to occur upon cellular exposure to non-ionic surfactants, [10][11][12] however little is understood about the underlying molecular mechanisms mediating these outcomes -the aspect studied in the present work.
The non-ionic, amphipathic surfactant studied here is composed of polyoxyethylene esters of 12hydroxystearic acid (marketed as Kolliphor® HS15, previously known as Solutol® HS15), and is used in various pharmaceutical applications including formulation of emulsions, 13,14 nanoparticles, 15,16 and as the primary component of permeability enhancing formulations. [17][18][19] It has a critical micellar concentration (CMC) between 0.06 -0.1 mM, a concentration range above which it was shown to enhance bioavailability of co-administered biotherapeutic compounds across epithelial cell barrier. 8,17,18,20,21 This effect has been attributed to the surfactant's ability to increase plasma membrane fluidity, as reported for other non-ionic surfactants and indeed other permeability enhancer compounds. 22,23 Furthermore, subsequent effects on cytoskeletal and junctional protein elements have been oberserved. 18,21 However, the effects of an increase in plasma membrane fluidity on other cell processes and cell health have not been elucidated in these studies. The present report hence focuses on deciphering the time-dependent succession of cellular events occurring in Caco-2 epithelial cells in response to their exposure to a range of the surfactant concentrations below and above the surfactant CMC. The elucidation of the mechanism of nonionic surfactant toxicity on epithelial cells will provide vital information to enable defining exposure conditions for safe application of such compounds and aid future development of drug delivery formulations.

Cell culture
Caco-2 human colonic cancer cells were obtained from the American Type Culture Collection

LDH release assay
The lactate dehydrogenase (LDH) assay was performed according to the manufacturer's instructions (Sigma Aldrich, TOX7 kit). 75 µl per well of supernatant sample was removed from treated cells and transferred to a 96-well plate. 150 µl per well of LDH reagent was then added and the resulting mix incubated for 25 min at room temperature in the dark. Absorbance was then measured at 492 nm. Relative LDH release was calculated with the absorbance at 492 nm for untreated control cells taken as 0%, and the positive control, 1.0 % TX-100, assumed to cause total cell lysis, as 100%. This concentration of TX-100 was determined to be capable of inducing total cell lysis at exposures ≥ 5 minutes (Supplementary Figure S1).

Cellular internalization of FITC-dextran 4000 Da (FD4)
To assess the permeability of the plasma membrane to cell impermeable solutes, FITC-Dextran 4000 Da (FD4; Sigma-Aldrich) was employed as a model permeant. 25,26 Solutions containing surfactant at specified concentration and FD4 (200 µg/ml) were applied to cells for 5-240 minutes.
Following exposure, cells were washed three times with PBS buffer and fluorescence measured at 490/525 nm (excitation/emission) to detect the presence of internalized FD4.

MTS assay
The cellular reduction of MTS reagent (CellTiter 96 Aqueous Cell Proliferation Assay, Promega) was assessed following surfactant exposure for predetermined times (5-240 minutes). Cells were incubated with 120 µl MTS solution (17 % v/v applied in EMEM) for 120 minutes, after which absorbance was measured at 492 nm. MTS reduction data was normalized by setting values of the untreated control cells as 1, and values from cells treated with 1% TX-100 as 0.

JC-1 assay
Changes in mitochondrial membrane potential were monitored by a JC-1 (Biotium) aggregation assay. 27 Following cells exposure to the surfactant solutions under different conditions, these were removed and cells washed twice with PBS buffer, prior to incubation with 50 µl JC-1 dye (5 µg/ml in EMEM) for 15 min at 37°C. Following removal of dye solution, wells were washed with PBS buffer prior to measuring fluorescence at 550/600 nm (excitation/emission) for detection of JC-1aggregates, and 485/535 nm (excitation/emission) for detection of JC-1 monomers. Jaggregate:monomer ratios were then normalized to values induced by the untreated control (set to a value of 1) and 1 µM valinomycin (Sigma-Aldrich) (set to a value of 0) was employed as a known depolarizing agent 28 (Supplementary Figure S2).

CellTox green assay
Integrity of the nuclear membrane was measured by the binding of CellTox Green (Promega) to nuclear DNA. 29

Hoechst 33342 /propidium iodide microscopy
Integrity of the nuclear membrane and nuclear fragmentation was measured by propidium iodide (PI; Thermo Fisher Scientific) uptake. 30

Determination of ROS induction
Intracellular ROS levels were assessed using the CM-H2DCFDA probe (Thermo Fisher Scientific). 31 After exposure to treatments, cells were loaded with 10 µM CM-H2DCFDA in HBSS for 30 minutes at 37°C. The probe was removed, cells washed, and fluorescent intensity measured at 492/520 nm (excitation /emission). Measured values were then normalized to the untreated control (set as a value of 1).

Detection of activated caspase-3/7
The CellEvent ® caspase-3/7 green detection reagent (Thermo Fisher Scientific) was employed to evaluate levels of activated caspase-3 or 7. 32 After exposure to surfactant solutions, 150 µl 2% (v/v) CellEvent probe in PBS was applied per well for 30 minutes at 37°C. Fluorescent intensity was measured at 502/530 nm excitation /emission) and normalized to the untreated control (set as a value of 1).

Calcium imaging
Intracellular Ca 2+ was monitored by epifluorescent microscopy with the Ca 2+ fluophore, FLUO-4 AM (Thermo Fisher Scientific). 33 Caco-2 cells were seeded on 15 mm borosilicate coverslips at a density of 1. For each visual field studied, regions of interest (ROI) were drawn around single and cluster of cells. These were corrected for background fluorescence by subtraction, and fluorescence intensity traces generated as a function over time. Image analysis was performed with custom scripts written in Labtalk (OriginLab Corporation, MA USA).

Statistical analysis
Concentration-response relationships were quantified by fitting the data with the equation: where R is the response magnitude, Rmin is the minimum value, Rmax is the maximum value, h is the slope index, [S] is the surfactant concentration and EC50 the concentration that produces halfmaximal effect.
Statistical analysis, unless otherwise stated, was performed by one-way ANOVA with Dunnett's multiple comparison post hoc test using GraphPad Prism (version 7.0). In the figures, statistical probability is indicated by: *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001. With multiple plots the statistical significance placed close to the relevant plot, and written in a coordinated colour to the plot. All data are presented as mean ± S.D from triplicates of three independent experiments, unless stated otherwise.

Surfactant exposure effect on Caco-2 cell membrane fluidity
Initially the effect of cells exposure to the non-ionic surfactant on plasma membrane fluidity was assessed using Laurdan general polarisation (GP) (Figure 1)     Data are mean ± S.D (N = 3, n = 3) and are displayed as the relative response to that observed with control (described in materials and methods). Statistical inference between various times for each assay are by 1-way ANOVA with Dunnett's multiple comparison test.

Surfactant effect on mitochondrial and cell metabolism
The mitochondrial effects of surfactant exposure are shown in Figure 4. As observed in Figure   . Effect of Kolliphor HS15 on Caco-2 mitochondrial and metabolic activity. Cells were exposed for 5-240 minutes to increasing concentrations of Kolliphor HS15 solutions as indicated. Mitochondria membrane potential was evaluated using the JC-1 assay, metabolic activity by MTS reduction and intracellular ROS levels assessed using the CM-H2DCFDA probe. Data are mean ± S.D (N = 3, n = 3) and are displayed normalized to that observed with control (described in materials and methods). Statistical inference between various times for each assay are by 1-way ANOVA with Dunnett's multiple comparison test.

Post exposure metabolic-decline and activation of effector caspases
To explore the reversibility of cellular effects to surfactant exposure, particularly reversibility of effects of short exposures (5-20 minutes), exposed cells were subjected to a 'recovery' post-  activation in Caco-2 cells. Cells were initially exposed to Kolliphor HS15 for 5, 10, 20 and 60 minutes and subjected to a 'recovery' period of 360 minutes in the absence of surfactant followed by the assessment of caspase activation. Data are mean ± SD (N = 3, n = 3) and are displayed normalized to that observed with control (described in materials and methods). Statistical inference between various times at any given concentration are by 1-way ANOVA with Dunnett's multiple comparison test.  Figure 6; bar chart). Figure 6 corroborates results from the DNA binding assay in Figure 3. λem 447/60 nm) and images merged using ImageJ software. Images shown are representative of 3 sets of independent images. Cells were exposed to treatments for 240 minutes and imaged on EVOS microscope 40X magnification (scale bar 100 µm). Bottom row, bar chart illustrates the mean ± S.D of nuclei size (µm 2 ) of treated cells (Neg, negative control; Kol., 10 mM Kolliphor HS15; Apop., 10 µM Staurosporine; Nec., Ethanol) measured using ImageJ analysis software and counting of at least 100 cells per group. Statistical significance measured using one-way ANOVA. Inserts (i, ii and iii) represent enlarged areas highlighted by white dotted boxes from Ho/PI micrographs and highlight morphological differences of nuclei between treatments. White arrows indicate nuclei with apoptotic features.

Effect of mitochondrial hyperpolarization on surfactant responses
To investigate the importance of early mitochondrial hyperpolarisation, cells were co-treated with the surfactant and a sub-toxic concentration of FCCP (0.5 µM), a mitochondrial protonophore known to depolarize mitochondrial membrane at high (toxic) concentrations and inhibit its hyperpolarization at low concentrations. 38,39 The applied FCCP concentration (0.5 µM) did not induce mitochondrial membrane depolarization, yet was able to diminish the increase in ΔΨm  Statistical differences between Kolliphor HS15 alone and Kolliphor HS15 co-treatment with FCCP are by 2-way ANOVA with Sidak's multiple comparison test.

Discussion
The increased permeability of epithelial cell layer to different permeants (e.g. small drug molecules and biologics) induced by non-ionic surfactants used as permeability enhancers has been connected to their effects on cell plasma membranes. 2,40 However, studies to decipher the mechanism(s) by which increased membrane permeability may be connected to other cellular events, including cytotoxicity of non-ionic surfactants are scarce. 3 In the present work we corroborate the time-dependent membrane, mitochondrial, and cell death-associated impacts of a non-ionic surfactant to develop a toxicity profile that will aid in the understanding of how these compounds can be safely utilized. Evaluating surfactant effects on cells as a function of exposure time, rather than typical measurements at a single time point, allowed us to discuss the surfactant effects in the context of time (as illustrated in Figure 8) and enable mechanistic information to be elucidated.

Figure 8. Suggested mechanism of toxicity of high concentrations of non-ionic surfactant on Caco-2 intestinal epithelial cells.
Exposure to surfactant includes rapid fluidization of the cell plasma membrane which is associated with minor fluctuations in intracellular Ca 2+ levels. This could incite a membrane-derived stress response, akin to the heat shock response, and is characterised by an early 'survival' phase (in red) and the subsequent initiation of a stress-induced mitochondrial-mediated apoptotic pathway (in blue). Occurring in parallel to these responses are the direct membrane effects of surfactant (in yellow) which, over time, progress from stressinducing disruptions in membrane fluidity, to membrane perturbations associated with increased membrane permeability. LDH, lactose dehydrogenase; FD4, FITC-dextran (4kDa); ΔΨm, mitochondrial membrane potential; perm., permeabilization.

Initial, primary surfactant effects
Changes in the fluorescence profile of the Laurdan probe (Figure 1) indicate the creation of an increasingly hydrophilic, fluid environment within the plasma membrane that occurs immediately upon surfactant application (within 0-1 minute), indicating fast initial incorporation of surfactant molecules into the plasma membrane structure. This initial effect does not appear to induce an immediate increase in membrane permeability, or impairment of plasma membrane barrier function, as judged from the initial absence of FD4 internalization and LDH release (Figure 3b).
The transient increase (~10%) in intracellular calcium levels within initial 3 minutes of exposure ( Figure 2) is most likely associated with plasma membrane fluidization and associated influx of extracellular calcium (Figure 8). 41 The initial level of cell membrane fluidization induced by the surfactant (10 mM) within 1 minute of exposure appears comparable in magnitude to that of membrane fluidization of cells exposed to 42°C heat shock (Figure 1e). This similarity suggests that the cells membrane fluidization by the surfactant exposure may be capable of triggering response akin to a mitochondrial heat shock response ( Figure 8). 41 In Caco-2 cells, exposure to 42 o C conditions employed in the current study has been previously demonstrated by others to induce heat shock response and expression of associated proteins. 35,36 It should be noted however that in cells exposed to a 42 o C heat shock the Laurdan fluorescence gradually returned to the initial level over time (Figure 1d), unlike in surfactant exposed counterparts. This apparent recovery of Laurdan GP was unexpected as the

Stress-induced survival response
Incorporation of surfactant molecules into, and consequent fluidization of the plasma membrane, have been demonstrated to induce clustering of membrane raft regions; 48 a phenomenon which alters the spatial coordination of membrane proteins involved in cellular thermal sensing and triggers the cellular heat shock response. 49 In addition to this membrane 'restructuring' on surfactant exposure and increased membrane fluidity (Figure 1), the observed transient increase in intracellular calcium ( Figure 2) may also play a role in activating the heat shock response; a phenomenon previously observed by others ( Figure 8). 41,50,51 The heat shock response has traditionally been attributed to protein denaturation, 52 however it is now widely established that cellular sensing and responding to stress signals occurs via the induction of membrane-associated signalling pathways. [53][54][55] Due to the nature of its structural molecules, membrane lipid bilayer organization is sensitive to temperature and confers the plasma as caspase inhibition has been recognized as a part of an anti-apoptotic environment and cell survival mechanisms, 59 including a survival response to heat shock ( Figure 8). 60,61 In addition, certain heat shock proteins are capable of binding to, and stabilizing heat shock perturbed membranes. 43 This membrane 'stabilization' may be protecting against further surfactant-induced membrane fluidization (between 1-30 minutes, Figure 1) and regulating calcium flux during early exposure times ( Figure 2). 62,63 In a similar manner, elements of the heat shock response may be responsible for the decrease in FD4 internalization observed at 5 and 10 minutes (Figure 3), which is further supported by a study conducted by Szöllősi et al. who report a decrease in FITC-dextran internalization in response to heat stress. 64 Thus the activation of the heat shock response may aim / attempt to protect the cell from early lysis by 'consolidating' the membrane and may limit surfactant-induced increase in membrane permeability at early time points by upregulating the membrane stabilizing proteins. 42,43,62,65

Metabolic effects
The reduction of the MTS tetrazolium salt to a formazan salt is mediated via NADH-dependent enzymes. 66  to address the stress stimuli. 70 Mitochondrial hyperpolarization has been suggested to be an early, key signalling element in the heat shock response. 41 Moreover, mitochondrial hyperpolarization is recognized as an early step in apoptosis. 39,71 In the present study, it appears that the induction of subsequent effector caspases via mitochondrial hyperpolarization is a time-dependent process; the induction of hyperpolarized mitochondria alone does not appear to trigger cell death (0-5 minutes), however its prolonged presence (>5 minutes) does ( Figure 8). In support of this idea, our data demonstrate that the inhibition of surfactant induced membrane hyperpolarization by FCCP prevents the activation of caspase 3/7 (Figure 7c).

Stress-induced apoptotic cell death
Our data reveal that exposure to surfactant concentrations above the CMC induce apoptosis, as indicated by effector caspases activation ( Figure 5b) and nuclear fragmentation ( Figure 6). In addition, unlike the necrotic control (ethanol), 72 surfactant exposure did not induce observable nuclear swelling ( Figure 6). These attributes would point to an induction of a form of apoptotic cell death, whereby the observed increases in cell plasma and nuclear membrane permeabilization ( Figure 3) are a 'direct' consequences of surfactant action, as opposed to being associated with necrotic processes (Figure 8).
Time course, 'recovery' experiments reveal that the activation of apoptosis appears to occur between 5-10 minutes of surfactant exposure at concentrations ≥ 1.0 mM ( Figure 5). Activation of the apoptosis appears to inhibit the cell survival response, as suggested by the sustained metabolic burst that occurs in the absence of mitochondrial hyperpolarization in the presence of FCCP ( Figure 7b); highlighting the potential crosstalk between these pathways. 73 This could be supported by high levels of MTS reduction that continued for almost 180 minutes into the post-exposure 'recovery' (Figure 5a), rather than this metabolic burst subsiding, as seen within 20 minutes in the presence of the surfactant (Fig 4).
The inhibition of apoptosis with FCCP did not however completely prevent mitochondrial membrane potential depolarization, as it was still observed after 180 minutes in the presence of FCCP (Figure 7a). This later depolarization is most likely the consequence of direct surfactant permeabilization of the mitochondrial membranes. Moreover, nuclear membrane permeabilization, which is also evident after ~120 minutes of exposure, was unresponsive to FCCP treatment (Supplementary Figure 10), a feature that indicates permeabilization of this membrane is also occurring as a direct surfactant effect. Both of these phenomena suggest the presence of surfactant molecules in intracellular organelle membranes after prolonged exposure.
How the surfactant molecules access membranes of intracellular organelles remains to be ascertained. Given the amphipathic nature of non-ionic surfactants, their molecules would, according to current understanding, at least initially ( Figure 1) accumulate in the plasma membrane. This insertion would (initially) occur into the outer leaflet of the membrane bilayer and, for non-ionic surfactants, the head group appears to strongly influence (if/when) a subsequent trans-bilayer diffusion occurs; estimated half-times ranging from 350 ms to several hours for e.g.
octaethylene glycol monododecyl ether and dodecylmaltoside, respectively. 46,74 Beyond plasma membrane incorporation (and effects on its fluidity), a possible contribution to surfactant molecules reaching intracellular membranes could be via the normal internalization of damaged membrane sections containing incorporated surfactant by plasma membrane repair mechanisms, as has been observed for membrane injury induced by pore-forming toxins or mechanical force. 75 A recent study illustrates accumulation of the surfactant molecules in the cell interior prior to its lysis, although it should be noticed that the work was conducted using ionic surfactant (sodium dodecyl sulphate). 76 The study proposes that intracellular membrane trafficking contributes to the surfactant uptake mechanism. Non-ionic surfactants have also been demonstrated to interact with, and form channels through lipid membranes in vitro. 77 As a consequence, surfactant molecules may diffuse though created pores (Figure 8), an event that would occur at latter times and higher surfactant concentrations, as potentially indicated by later influx of FD4 and LDH leakage ( Figure   3).
Finally, one could view apoptotic cell death as actually conferring protection to the cell population, or tissue as a whole, from surfactant-induced immunogenicity. The loss of cell membrane barrier function, and the consequent leakage of intracellular components, will likely promote immunogenic response in the surrounding cells and tissue. 78,79 The induction of rapid mitochondrially-mediated apoptosis can thus be viewed as advantageous, as it might minimize the toxic potential of surfactant exposure.

Conclusions
The primary observation of this study is that cell membrane fluidization caused by exposure to non-ionic surfactant is a process akin to thermal stress, resulting in the cellular heat shock response.
This has a relevance to the behaviour of other non-ionic surfactants used in therapeutic formulations, such as the alkylglycosides 2,44 and polysorbates, 45 which have been suggested to mediate their increases in cell membrane permeability via the induction of membrane fluidization.
Taken together, our data suggest that the safe use of non-ionic surfactants which operate by such a mechanism may be limited by the fact that their membrane permeability action is intrinsically linked to alterations in membrane fluidity and, an induction of apoptotic cell death. The work performed here therefore provides a foundation from which the study of other non-ionic surfactants could be furthered. Similarly, the findings reported here on intestinal epithelium may be applicable to permeability studies performed on other epithelial layers, such as airway epithelium.
Our study indicates that non-ionic surfactant cytotoxicity is induced by membrane effects, however, it is the mitochondrial function and mitochondria associated responses, that are consequently triggered, that in fact mediate the majority of observed cytotoxicity, not membrane and Kolliphor HS15 exposure on ROS and CellTox Green assay results.